The performance of two adaptive robot control algorithms, namely the a
daptive algorithm by Slotine and Li and the proposed direct adaptive f
uzzy control algorithm, were studied. In Slotine and Li's method, the
algorithm consists of proportional derivative feedback and a full dyna
mics feedforward part, with the unknown manipulator and payload parame
ters being estimated on-line. The algorithm depends heavily on the par
ticular structure of manipulator dynamics and needed three simplifying
properties to derive the adaptive law. The direct adaptive fuzzy algo
rithm uses a fuzzy logic system as the controller. The derivation of t
he controller does not require any knowledge of the manipulator dynami
cs and the uncertainty in the system, only the boundary of the dynamic
s. The predominant concern of the two adaptation laws is to reduce the
tracking errors. In particular, they require no feedback of joint acc
elerations. Theoretical results and simulation studies on a two-link r
obot manipulator show that both methods are robust and stable. The dir
ect adaptive fuzzy controller outperforms the adaptive controller in m
ass variation. (C) 1996 John Wiley and Sons, Inc.